Hyperparameter Tuning with Logistic Regression

Optimisation and Deep Learning course project

Kaggle Dataset: https://www.kaggle.com/geomack/spotifyclassification

Contents

1. Data Understanding
2. EDA
3. Data Transformation: Feature Scaling using RobustScalar
4. Feature Importance: Tree models - Decision Tree, Random Forest and Gradient Boosting
5. Modelling: Logistic Regression
6. Tuning: GridSearchCV & RandomizedSearchCV

Findings

• There is no strong correlation between likeability of the songs vs. the songs attributes present in the dataset chosen
• The data was then tuned using the aforementioned tuning methods including data splitting (80 - 20) and choosing the features shown during Feature importance.